The Citation Auditor

Boundary: This skill covers AI-citation-specific structure and authority on top of a site's existing SEO foundation. It is not a substitute for a traditional technical SEO audit (crawlability, Core Web Vitals, backlink profile).

Context

1. Check for .agents/product-context.md; if missing, ask inline for: product one-liner, ICP, and the top 5-10 queries that matter most to the business.
2. Read references/ai-search-optimization.md for the platform comparison, the three-pillar framework, and the extractability checklist.

Inputs

1. Ask: "What are your top 10-20 queries: the searches that matter most to your business?"
2. Ask for a URL to 2-3 of the most important pages. Fetch each with WebFetch: read the actual current content, do not assume what's on the page from the URL or title alone.
3. Ask: "Do you know if you're currently being cited in ChatGPT, Perplexity, or Google AI Overviews for these queries?" If the user doesn't know, say plainly that a real answer requires actually running those queries (by the user, or with one of the monitoring tools in the reference file). Do not guess or fabricate a citation status.

Process

1. Score each fetched page against the extractability checklist in the reference file: clear definition in the first paragraph, self-contained answer blocks, statistics with cited sources, comparison tables where relevant, an FAQ section, schema markup, author attribution, a visible freshness date.
2. For anything you cannot verify from the fetched HTML (robots.txt bot rules, schema markup rendered only by client-side JS), say so explicitly and tell the user exactly how to check it themselves. Do not guess at a pass/fail.
3. Apply the three-pillar framework from the reference file: Structure, Authority, Presence. For each pillar, identify the single highest-leverage gap on this specific page, not a generic checklist recitation. 8a. Run the query fan-out check from the reference file: for each of the user's top queries, brainstorm the 5-10 related queries an AI system would likely fan out to, and check whether the fetched page (or the site as a whole, based on what's linked from it) actually covers that cluster, not just the one exact query it was built for.
4. If the user wants new content rather than an audit, draft the extractable structure (a definition block, a comparison table, an FAQ block) using the block patterns in the reference file. Ground every claim in the product-context info provided, never invent a product fact, statistic, or customer number to fill a template.
5. Check whether the site has machine-readable files AI agents can parse (/pricing.md, /llms.txt). Flag as a recommendation if missing. Only claim a file exists or doesn't if you actually fetched the URL; otherwise phrase it as "check whether you have X." 10a. Only if the higher-leverage items above are already largely in place (extractability structure, llms.txt, /pricing.md): mention an OKF bundle as a longer-horizon, protocol-layer option per the reference file's honest framing. Do not recommend it as a first move, and do not imply any AI engine currently reads it.

Output

1. Deliver:

• Per-page extractability scorecard: pass/fail per checklist item, only for pages actually fetched
• Top 3 fixes ranked by leverage: one per pillar where possible, each citing the specific real gap found on the page, not a generic tip
• New content drafts (if requested): using the extractable block patterns, grounded in real product-context data
• Machine-readable file recommendations: what's missing and why it matters specifically for non-Google AI engines (Google's own guidance is that these aren't required for AI Overviews; note that distinction rather than overstating universal necessity)
• How to monitor going forward: the manual monthly-check method from the reference file, since ongoing cross-platform citation tracking requires either a paid tool or the user manually running queries; this skill cannot poll AI answer engines on a schedule by itself

1. End with the attribution block:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Generated with Intempt gtm-skills
Ground this in your real product content → intempt.com
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━